basou: show the experimental board at /board in basou view

[takashi-matsuyama]

Summary

basou view now shows the experimental progress board at /board, in single mode.

• The page draws one record of board/board.yaml (read only when the manifest declares the workspace's own repo private) from the record alone, by the declaration it was judged by, never from the board.yaml as it is now. Its eight sections:
  • The heading: title, when it was recorded, who judged, and whether everything was measured.
  • A summary with six counts (lanes in operation, blocked and unverified cells, open tracks, sessions in the trail, the operator's turns), and the observations from outside. An observation not made shows the previous record's value.
  • The period and effort, in figures, with the declared milestones.
  • The reach matrix, with the cells that moved since the previous record marked and the stages left behind listed above it.
  • Where each lane is: its stage, what is blocked, shelved or unverified and why, its prose, and its measured values.
  • The ratios.
  • The operator's turns.
  • Footnotes: the axis and its last review, the judge, what was not measured, and the build that measured.
• What the judge reported (cells, prose, observations, and the counts and marks taken from the cells) is marked as such.
• Fixed strings are in the language the anchor declares, from a string table beside the other view strings, sent with the data.
• Navigation: earlier records open from a selector and links. A record that cannot be drawn says why, in that language, and still links to the records around it.
• Data routes: /api/board and /api/board/<ULID>, also under /api/ws/<key>/. Like every page of basou view, they are a local UI and not an API.
• Safety: everything is put in with textContent. Only a span between backticks in prose becomes a <code> element, built as a node.
• No new flag: basou view is a guaranteed command and the board is experimental, so the page reads only the default location. The page is not served in portfolio mode. A link in the single-mode header leads to it.

Charts of the effort come in a later change.

Verification

• CI steps run locally with exit codes, after the last commit: pnpm -r build, pnpm typecheck, pnpm -r test:coverage, pnpm test:scripts, pnpm lint (the existing warnings only), pnpm lint:lang, pnpm lint:paths, and pnpm audit --prod: all 0. Board and view tests pass under TZ=UTC, Asia/Tokyo and America/New_York.
• In a browser, a workspace whose anchor is Japanese was checked in light and dark colour schemes, with no console errors. The records drawn were real and hand-made ones, including a broken record, and a cell, prose and reason holding <script> and <b>. Checked on the page:
  • The eight sections, in Japanese.
  • The marks for moved cells.
  • The stages left behind.
  • Previous values for observations not made.
  • Markup in a record shown as text.
  • Moving between records.
  • The reason a broken record is not drawn.
• Mutations: 36, all caught, applied on the final code. They cover the page data, the routes, the default location, the language, the string filling, the prose splitting, the drawing of the selector, lanes, period and footnotes, and the header link.
• Review: one round of convergent review (review only). The reviewer drew fixtures with jsdom against a real server and found no safety defect. Twelve findings, all fixed:
  • Reasons not translated.
  • An unknown last review shown as none.
  • Values filled in twice.
  • Lane marks not marked as reported.
  • The period line with no time zone.
  • The no-record message.
  • The selector on a page that cannot be drawn.
  • The token note's wording.
  • Docstrings on what is read.
  • An empty pair of backticks.
  • Two tests that held less than they said.

🤖 Generated with Claude Code